High-Temperature Thermochemical Energy Storage Based on Redox Reactions Using Co-Fe and Mn-Fe Mixed Metal Oxides Laurie André, Stéphane Abanades, Laurent Cassayre

High-temperature thermochemical energy storage based on redox reactions using Co-Fe and Mn-Fe mixed metal oxides Laurie André, Stéphane Abanades, Laurent Cassayre

To cite this version:

Laurie André, Stéphane Abanades, Laurent Cassayre. High-temperature thermochemical energy storage based on redox reactions using Co-Fe and Mn-Fe mixed metal oxides. Journal of Solid State Chemistry, Elsevier, 2017, 253, pp.6 - 14. 10.1016/j.jssc.2017.05.015. hal-01786111

HAL Id: hal-01786111 https://hal.archives-ouvertes.fr/hal-01786111 Submitted on 20 Jul 2018

HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non, lished or not. The documents may come from émanant des établissements d’enseignement et de teaching and research institutions in France or recherche français ou étrangers, des laboratoires abroad, or from public or private research centers. publics ou privés. Open Archive Toulouse Archive Ouverte

OATAO is an open access repository that collects the work of Toulouse researchers and makes it freely available over the web where possible

This is an author’s version published in: http://oatao.univ-toulouse.fr/20472

Official URL : http://doi.org/10.1016/j.jssc.2017.05.015

To cite this version: André, Laurie and Abanades, Stéphane and Cassayre, Laurent High-temperature thermochemical energy storage based on redox reactions using Co-Fe and Mn-Fe mixed metal oxides. (2017) Journal of Solid State Chemistry, 253. 6-14. ISSN 0022-4596

Any correspondence concerning this service should be sent to the repository administrator: [email protected] High-temperature thermochemical energy storage based on redox reactions using Co-Fe and Mn-Fe mixed metal oxides

Laurie Andréa, Stéphane Abanades a,⁎, Laurent Cassayre b a Processes, Materials, and Solar Energy Laboratory, PROMES-CNRS, 7 Rue du Four Solaire, 66120 Font-Romeu, France b Laboratoire de Génie Chimique, Université de Toulouse, CNRS, INPT, UPS, Toulouse, France

ABSTRACT

Keywords: Metal oxides are potential materials for thermochemical heat storage via reversible endothermal/exothermal Metal oxide redox reactions, and among them, cobalt oxide and manganese oxide are attracting attention. The synthesis of Thermal heat storage mixed oxides is considered as a way to answer the drawbacks of pure metal oxides, such as slow reaction Solid-gas reaction kinetics, loss-in-capacity over cycles or sintering issues, and the materials potential for thermochemical heat Reduction storage application needs to be assessed. This work proposes a study combining thermodynamic calculations Oxidation and experimental measurements by simultaneous thermogravimetric analysis and calorimetry, in order to Thermodynamic equilibrium calculations identify the impact of iron oxide addition to Co and Mn-based oxides. Fe addition decreased the redox activity

and energy storage capacity of Co 3O4/CoO, whereas the reaction rate, reversibility and cycling stability of Mn 2O3/Mn3O4 was signiﬁcantly enhanced with added Fe amounts above ~15 mol%, and the energy storage capacity was slightly improved. The formation of a reactive cubic spinel explained the improved re-oxidation yield of Mn-based oxides that could be cycled between bixbyite and cubic spinel phases, whereas a low reactive tetragonal spinel phase showing poor re-oxidation was formed below 15 mol% Fe. Thermodynamic equilibrium calculations predict accurately the behavior of both systems. The possibility to identify other suitable mixed oxides becomes conceivable, by enabling the selection of transition metal additives for tuning the redox properties of mixed metal oxides destined for thermochemical energy storage applications.

1. Introduction possible. The reaction enthalpy is stored in the reaction products during the heat charge (Eq. (1)), and can be released by reversing the The purpose of thermal energy storage (TES) in solar thermal reaction during the discharge (Eq. (2)). power plants is to store solar heat in the form of sensible, latent or MO + Heat → MO + α O (1) chemical energy during on-sun hours and then to use it during o ff-sun (ox) (red) 2(g) hours in order to match a variable electricity demand with an MO + α O → MO + Heat (2) intermittent energy source supply, thus enabling energy production (red) 2(g) (ox) according to needs and enhancing energy generation dispatchability An advantage of using metal oxides as TES materials is the use of [1]. This study focuses on thermochemical heat storage, which shows air as the heat transfer fluid that can be processed in an open-loop advantages over latent and sensible heat storage, including higher system. Such thermochemical systems are suitable for application in energy storage densities, possible heat storage at room temperature in high-temperature solar power plants based on central air receiver the form of stable solid materials, and long term storage in a broad (pressurized air-based solar tower receivers for power generation via – temperature range (400 1200 °C) with heat released at a constant gas turbines). Cobalt oxide (Co3O4) and manganese oxide (Mn 2O3), fi restitution temperature de ned by the reaction equilibrium. which reduce into CoO and Mn 3O4 respectively, are among the most Concentrated solar power (CSP) provides the heat source required studied metal oxides considered as promising materials for thermo- for endothermal/exothermal reversible reactions involved in thermochemical energy storage. Experiments showed that Co 3O4 is the most chemical energy storage (Eqs. (1) and (2)) in order to store solar energy suited raw material given the fast reaction kinetics, complete reaction in the form of chemical bonds. When combined with CSP systems, the reversibility and cycling stability with a measured gravimetric energy continuous production of electricity in thermal power plants becomes storage density of 576 kJ/kg [2] (while reported theoretical enthalpy is

⁎ Corresponding author. E-mail address: [email protected] (S. Abanades). Fig. 1. XRD analysis of fresh synthetic Co 3O4 with (a) 40 mol% Fe, (b) 25 mol% Fe, (c) 10 mol% Fe, and (d) 5 mol% Fe.

844 kJ/kg [3 –6] ). However, the cost and potential toxicity of cobalt of the material. Furthermore, the addition of iron helped to stabilize oxide may require the development of other alternative materials. The and enhance the oxidation rate of manganese oxide over thirty redox reduction step of Mn 2O3 was observed in the range of 920 –1000 °C, cycles. According to their study, the fastest and most stable oxidation with notably slow re-oxidation in the range of 850 –500 °C, with a reaction was obtained for Mn 2O3 doped with 20 mol% Fe. In another reported gravimetric energy storage density of 110 –160 kJ/kg [7,8] study, the authors also considered Fe-Cu co-doping in manganese and a theoretical enthalpy of 202 kJ/kg [8]. The re-oxidation usually oxide and showed that the incorporation of Cu diminished the happens in two steps, the first one being during the cooling and in reduction temperature, whereas the incorporation of Fe increased the between 700 °C and 500 °C and the second one being during the re- oxidation temperature [12]. heating and in the range of 500 –850 °C [9] because of strong kinetic Recent models provide thermodynamic descriptions of the Co-Fe-O limitations. In order to complete the slow re-oxidation during the [13,14] and Mn-Fe-O [15,16] systems, based on exhaustive studies of cooling step, it needs to last long enough by decreasing drastically the available experimental data. In the present work, some of these models cooling rate. were used to calculate the nature and composition of solid phases at Optimization of materials reactivity is generally required for metal equilibrium, the oxygen storage capacity (equal to the mass loss during oxide species by using e.g. doping strategies, controlled synthesis reduction), the transition temperature, and the reaction enthalpy of Co- techniques for tailored morphology, or stabilization with inert materi- Fe and Mn-Fe mixed oxides, with Fe content up to 50 mol%. als to alleviate sintering e ffects. Improvement of materials properties Equilibrium calculations were compared to experimental (thermogra- and flexibility such as reaction kinetics and reaction temperature vimetric and calorimetric) measurements carried out with various tuning can be obtained by the addition of dopants. In the speci fic case oxide compositions, under 20%O2/Ar atmosphere, in the 750 – of the Mn 2O3/Mn3O4 cycle, the doping with transition metals is an 1050 °C temperature range. The combination of experimental data option to relieve the slow re-oxidation issue [2,4]. and thermodynamics aims at providing a better understanding of the The present study aims at investigating the e ffect of Fe addition on behavior of the mixed oxides and de fining the optimal composition. the performances of Co and Mn metal oxides, which have been identified as interesting candidates for TES [10]. The optimal mixed metal oxide composition to reach the highest performances for TES is 2. Materials and methods addressed, encompassing high reaction enthalpy, stability upon cycling, low temperature gap between reduction and re-oxidation, and 2.1. Synthesis and characterization reaction in a temperature range adapted to CSP. The addition of transition metals is investigated as a way to answer the drawbacks of Pure Co 3O4 and cobalt oxide mixed with 5, 10, 25, and 40 mol% Fe simple metal oxides, including slow reaction kinetics [5], sintering as well as pure Mn 2O3 and manganese oxide mixed with 10, 15, 20, 30, [11], loss-in-capacity over cycles [8], etc. Block et al. [2] reported that 40 and 50 mol% Fe were synthesized. In the following, the oxides both addition of iron to cobalt oxide or addition of cobalt to iron oxide compositions are either noted x(Fe) or mol% Fe, referring to the molar results in lower enthalpies of reaction compared to those of the pure content of Fe on metal basis (based on a Fe/(Fe+Co) and Fe/(Fe+Mn) fi oxides. However, they estimated that cobalt oxide doped with 10% of mole ratio). All the powders were synthesized via a modi ed Pechini iron possesses higher reduction/oxidation reversibility than pure method [2], using metal nitrates ( > 98% purity), citric acid ( > 99% cobalt oxide and still shows high enthalpy of reaction. Pagkoura et al. purity) and ethylene glycol ( > 99% purity) in aqueous solution. The [9] showed that cobalt oxide with 10 –20 wt% of iron shows good solution was heated up until obtaining a viscous solution. The powders thermo-mechanical stability over ten redox cycles. Carrillo et al. [11] were then calcined in air at 200 °C for 2 h and at 750 °C for 4 h in order showed that the addition of iron does not allow avoiding the sintering to fully eliminate the organic fraction and residues of the synthesis, and to stabilize the structure. Higher calcination temperatures were not encountered with Mn 2O3 but helps to increase the heat storage density considered to avoid the reduction of the materials during the synthesis. Fig. 2. XRD analysis of fresh synthetic Mn 2O3 with (a) 50 mol% Fe, (b) 40 mol% Fe, (c) 30 mol% Fe, (d) 20 mol% Fe, and (e) 10 mol% Fe.

The powders were characterized by X-ray di ffraction (XRD) before heating rate was 20 °C/min for the reduction step, and the cooling rate being studied in redox cycles. XRD analysis was performed ( Figs. 1 and was 10 °C/min for the re-oxidation step. For some runs with cobalt 2, SI1, SI2) at room temperature using a PANalytical XPert Pro based oxides, the temperature was maintained for 15 min at 1050 °C di ffractometer (CuKα radiation, λ = 0.15418 nm). X-ray di ffraction and at 800 °C for achieving the state of reaction equilibrium. For all the measurements of θ-θ symmetrical scans were made over an angular materials, a series of three cycles was performed for each studied range of 10 –80°. The step size and the time per step were fixed at 0.01° formulation in order to determine their suitability for achieving and 20 s, respectively. The contribution from K α2 was removed and the reversible reactions and to assess a possible chemical deactivation of X-ray di ffractograms were recorded and studied using the PANalytical the materials as a result of a loss of redox stability during thermal software. The instrumental function was determined using a reference treatment. material (SRM 660, lanthanum hexaboride, LaB6 polycrystalline sample) and can be expressed by a polynomial function [17]. The synthesized oxide phases were corresponding to the structures of (i) a 2.3. Thermodynamic calculations cubic spinel solution (Co1−xFe x)3O4 (Fig. 1) and (ii) a bixbyite solution Equilibrium calculations were performed with the FactSage 7.0 (Mn1−xFe x)2O3 [18] (Fig. 2), with traces of iron oxide only detected for the materials containing the largest amount of iron ( Figs. 1a and b, and software [19] and the FToxid database, which includes models of the thermochemical properties of the oxide phases in Fe-Co-O [13] and Fe- 2a). Above 20 mol% Fe added to Co 3O4, the material is no longer a single spinel phase but a mixture of two spinels. Mn-O [15] systems. For calculations in the Fe-Co-O system, the following solid phases were considered: the Fe 2O3 compound, the The morphology of the synthesized (Mn 1−xFe x)2O3 phase was also characterized (Fig. SI3) by SEM (FESEM, HITACHI S4800). The very cubic spinel solution (Co1−xFe x)3O4 (C-Spin) and the monoxide solu- porous structure obtained with the Pechini method can be observed in tion (Co1−xFe x)O 1+y (Monoxide). For the Fe-Mn-O systems, four phases were considered: two spinel solutions (Mn1−xFe x)3O4, one with cubic Mn 2O3 samples synthesized with no iron ( Fig. SI3a and b). The porous structure is still present with the addition of iron during the synthesis; structure (C-Spin) and the other with tetragonal structure (T-Spin), the however, the porosity reduces with increasing iron content. The sample bixbyite solution (Mn1−xFe x)2O3 (Bixb) and the corundum solution (Fe1−xMn x)2O3 (Cor). The monoxide and spinel solid solutions were with 10 mol% Fe ( Fig. SI3c and d) still resembles the pure Mn 2O3 sample, while the sample with 50 mol% Fe appears di fferent (Fig. SI3e described within the framework of the Compound Energy Formalism and f), with thicker necks forming the pore walls. This change in [20], with two and three ionic sublattices, respectively (see [13,15] for morphology does not alter the reactivity, as evidenced in the experi- more details). The bixbyite and corundum phases were modelled as a mental section. random mixture of the two end-members (Fe 2O3 and Mn 2O3), with no interaction parameter [15]. Two gaseous components, O 2 and Ar, were taken into account, with ideal mixing properties. The total pressure was 2.2. Experimental procedure for thermochemical redox cycling always 1 atm. For both systems, two kinds of calculations were carried out: (i) plot Experimental data concerning transition temperatures during re- of phase diagrams, (ii) system equilibrium at various compositions and dox process, oxygen storage capacity and reaction enthalpies were temperatures, providing, for each composition and at each temperature obtained by coupled thermogravimetric analysis (TGA) and di fferential step, the phase assemblage and the cationic distribution in each scanning calorimetry (DSC), using a Netzsch STA 449 F3 System. solution phases, as well as the total enthalpy of the system. About 50 mg of powder was placed into an alumina crucible and then The second type of calculations gives access to the theoretical mass subjected to simultaneous thermal analysis (STA) consisting of TGA loss (i.e. oxygen storage capacity), Δm, as de fined by (Eq. (3)). This coupled with DSC. quantity is directly comparable to the mass change measured by TGA. Reduction-oxidation cycles were performed in a 20% O 2/Ar atmo- mass of solid at T− mass of solid at T ∆m (%) = 100. max min sphere (10 N mL/min O 2 and 40 N mL/min Ar), between 800 °C and mass of solid at T min (3) 1050 °C for cobalt based oxides and between 750 °C and 1050 °C for manganese based oxides. Each run also included a last reduction step where Tmin and Tmax are referring to the minimal and maximal under Ar (99.999% purity), in order to get information about the temperature of the thermochemical cycle. reduction temperature of the material under inert atmosphere. The The enthalpy of the reduction reaction (Eq. (1)), ΔrH, was calculated according to (Eq. (4)). This quantity is compared to the heat of reaction determined by DSC analysis. For reactions occurring on a large temperature range (see Section 3), this calculation includes a contribution due to the calori fic capacity of the compounds (sensible heat), which leads to an overestimation of the enthalpy strictly related to the reaction.

0 1 s1 s2 O2 ∆r H (kJ/kg of s ) = [mHTs1 ()+end mHTs2 ()+end mH O 2 ()] T end ms0 s0 −H ( T beg ) (4) where Tend is the temperature (°C) at the end of the phase transition or the maximal temperature of the cycle; T beg is the temperature (°C) at 0 the beginning of the phase transition; m S is the mass (kg) of the solid 1 2 phase at T min; m S and mS are the mass (kg) of the solid phases at x Tend; m O2 is the mass (kg) of dioxygen at T end; H (T) is the enthalpy (kJ/kg) of the compound x at the temperature T.

3. Results and discussion

3.1. Cobalt-based mixed oxides

The calculated Co-Fe-O phase diagram for pO 2 = 0.20 atm is presented in Fig. 3a. It shows that, at equilibrium, the iron content greatly in fluences the composition and the amount of the monoxide phase formed at 1050 °C when heating the spinel phase from 800 °C. For small Fe additions (0 < x(Fe) < 0.1), the spinel phase is fully converted into the monoxide phase. At very high Fe content (x(Fe) > 0.6), there is no phase transition achievable below 1200 °C, which is the upper temperature limit usually considered for TES applications. For intermediate Fe content (0.1 < x(Fe) < 0.6), an increasing proportion of spinel phase is not converted into monoxide. As an example, Fig. 3b presents the in fluence of the temperature on the phases amount for x(Fe) = 0.25. A temperature of about 1460 °C is required to fully convert the spinel phase into the monoxide phase. Furthermore, as illustrated in Fig. 3c for x(Fe) = 0.25, the monoxide phase (Co1−xFe x)O 1+y presents a noticeable over-stoichiometry (y) in oxygen, which increases with temperature. The TGAs of cobalt-based oxides are presented in Figs. 4 and 5 with di fferent temperature programs. A first observation is that a noticeable change in the re-oxidation onset temperature ( Fig. SI4) is measured between the samples with 0–10 mol% Fe ( Figs. 4b–c, 5c–e) and with 25 –40 mol% Fe ( Figs. 4a, 5a–b). Indeed, the re-oxidation step starts at the beginning of the temperature decrease (i.e. 1050 °C) for the samples containing 25 mol% Fe and 40 mol% Fe, which is in line with the phase diagram (Fig. 3a) since the two-phase equilibrium strongly depends on the temperature. The Co-Fe samples were kept in a reduced state after TGA (final cooling step in Ar to avoid re-oxidation in Fig. 5) and analysed via XRD (Fig. SI1). Phase identi fication confirmed the Fig. 3. (a) Calculated Co-Fe-O phase diagram at pO 2 = 0.20 atm, (b) calculated temperature evolution of the mass of solid phases for x(Fe) = 0.25, (c) calculated presence of the single monoxide phase for low Fe contents (0 and 5 mol temperature evolution of oxygen over-stoichiometry (y) in the monoxide phase

%), whereas the samples consisted of a mixture of monoxide and spinel (Co0.75Fe 0.25)O 1+y. phases for higher Fe contents (10 –40 mol%), in agreement with the calculated phase diagram. denced. Indeed, as both temperatures rise with Fe addition, it can be The in ﬂuence of Fe incorporation on reaction temperatures is noticed that the gap between the reduction and oxidation temperatures reported in Fig. 6 (Fig. SI4 compares data under Ar and 20%O 2/Ar reduces. For 40 mol% Fe, the onset temperature for reduction has atmosphere). Both temperatures at peak reaction rate (i.e. tempera- increased by 50 °C compared to pure Co 3O4, while the onset temperatures at which the reaction kinetics reached its maximum, Fig. 6) and ture for oxidation has increased by 90 °C. The reduction of this onset temperatures (Fig. SI4) were measured. The measured onset temperature gap is an asset for large-scale applications since it reduces temperature for reduction of Co 3O4 to CoO is about 920 °C, which is in the amount of energy spent for the heating and cooling of the system good accordance with previous studies carried out in air atmosphere, between the charge and the discharge steps. However, an increase in reporting temperatures ranging from 880 to 930 °C [6,7,9,21]. A the oxidation temperature may also be a disadvantage if the system gradual increase of the reduction and oxidation temperatures is must be externally heated to initiate the oxidation (in case the material observed with increasing amount of iron in the material, which is is stored in reduced form at room temperature). consistent with the phase diagram (Fig. 3a). An e ﬀect of Fe addition on In Fig. 7 are reported, along with the calculated equilibrium values, the temperature gap between reduction and oxidation is also evi- Fig. 7. Average and maximal experimental mass loss (Δm in %) of the Co-Fe mixed oxide between 800 °C and 1050 °C compared to Δm at thermodynamic equilibrium. Fig. 4. TGA of Co 3O4/CoO with addition of (a) 25 mol% Fe, (b) 10 mol% Fe, (c) 5 mol%

Fe, including a ﬁnal re-oxidation step under 20% O2/Ar at the end. implies that a full conversion has been reached.

According to the thermodynamic model, the decrease in O 2 storage capacity with increasing Fe content is due to a combination of two factors. First, as evidenced by the phase diagram and the temperature proﬁle for x(Fe) = 0.25 ( Fig. 3b), the upper temperature limit of 1050 °C does not lead to a full conversion of the spinel phase when x(Fe) > 0.10. This contribution is represented as C1 in Fig. 7. It does not account for the decrease in Δm evidenced for x(Fe) = 0.05 and 0.10 (C2 in Fig. 7), which should be closer to the mass loss calculated according to (Eq. (5)) since a full conversion into monoxide occurs. This second factor is linked with the fact that the monoxide solid solution becomes non stoichiometric with the incorporation of Fe (Fig. 3c). This leads to a residual storage of oxygen in the monoxide phase, according to (Eq. (6)):

(Co1−x Fe x34() ) O s = 3 (Co 1− x Fe x1+() )O ys + (1−3y)/2 O 2 (g ) (6)

Fig. 5. TGA of Co 3O4/CoO with addition of (a) 40 mol% Fe, (b) 25 mol% Fe, (c) 10 mol% where y is the over-stoichiometry of oxygen in the monoxide phase. Fe, (d) 5 mol% Fe, (e) no Fe added, ending on a ﬁnal reduction step under Ar. As illustrated in Fig. 7, the calculated contribution of the oxygen over-stoichiometry in the monoxide phase at 1050 °C accounts for a loss of storage capacity of about 1% per unit mass of spinel when x(Fe) > 0.10, as compared to a maximum capacity of 6.6% for pure cobalt oxide.

As for the cycling stability, for pure Co 3O4, the average Δm is the same as both the theoretical value and the maximum value measured

experimentally. This means that the amount of O 2 released and regained did not change, conﬁrming the good material cycling stability over three cycles, as already reported by several authors [2 –7] . For the samples containing 5 and 10 mol% iron, since the average Δm falls

slightly below the theoretical and maximum value, a minor loss in O 2 exchange capacity upon cycling is evidenced. Conversely, the amount of

O2 exchanged remains stable over cycles for larger Fe contents (x(Fe) > 0.10) at the expense of a decreased oxygen storage capacity. Thus, the cycling stability of cobalt oxide does not su ﬀer from the addition of ﬁ Fig. 6. Experimental temperatures at peak reaction rate for cobalt-based oxides in 20% iron, as the re-oxidation conversion rate is not signi cantly decreasing

O2/Ar and comparison with the calculated temperature of the C-Spin/Monoxide over multiple cycles for Co-Fe mixed oxides, similarly to pure Co 3O4. transition. An additional cycling test was performed with a Co-Fe mixed oxide (10 mol% Fe) subjected to multiple consecutive cycles (12 cycles) in order the maximum values of Δm measured for each oxide composition, to demonstrate the stability of the material regarding both oxygen which means the highest amount of O 2 released during reduction, and storage capacities and energy storage densities (Fig. SI5). This con firms the average Δm for the three redox cycles. The hypothetic case of a full the robustness of the active redox materials and the ability to cycle conversion of the initial spinel into a stoichiometric monoxide, without deactivation. calculated according to (Eq. (5)), is also plotted in Fig. 7. Regarding the effect of the gas flow composition, the measured Δm due to the reduction under pure Ar was not signi ficantly di fferent from (Co1−x Fe)O x34() s = 3(Co 1− xs Fe)O x() + 0.5O(g 2 ) (5) the one measured with 20% O2/80% Ar. However, as illustrated by Fig. Measurements of the Δm values clearly show that increasing the SI4, the reduction onset temperature under Ar was decreased by 60 – amount of iron in the material reduces the amount of O 2 it is able to 80 °C, to reach a value of about 860 °C, almost independent of x(Fe). release at a given temperature. Furthermore, the maximum Δm values These data are rather di fferent from the equilibrium state calculated are in excellent accordance with equilibrium calculations, which with the thermodynamic model (Fig. SI6). Indeed, under pure Ar, the decreasing with increasing amount of Fe: the highest measured

enthalpy is 597 kJ/kg for pure Co 3O4, while the lowest is 50.7 kJ/kg for x(Fe) = 0.40. It should be noted that the DSC peaks for x(Fe) = 0.40

are very small, which makes them di ﬃcult to analyze. For pure Co 3O4, our data is close to the value of 576 kJ/kg obtained by [2] with a similar experimental procedure. As recently discussed in details by Block et al. [22], this kind of DSC measurements, operated in an open reactor at a rather fast heating rate, is not able to reproduce the tabulated enthalpy value, which is commonly reported as 844 kJ/kg [3 –6] . To explain the large discrepancy, the authors state that the contribution of a Co 3+ spin-state change, associated with an enthalpy of about 222 kJ/kg, is not measured by dynamic techniques. The thermodynamic model used in this study [13], which takes into account the heat capacity anomaly due to spin-state transition of Co 3+ , leads to an enthalpy of reaction of 749 kJ/kg for x(Fe) = 0. This is noticeably lower than the commonly admitted value of 827 kJ/mol [2,22]. For higher Fe contents, the calculations reproduce well the

general trend (decrease of ΔrH) with a systematic gap of about 180 kJ/ Fig. 8. Temperature evolution of the amount (kg) of O (g) released per kg of cobalt- 2 kg. based spinel containing various amounts of Fe. In TES application, the enthalpy of the reaction is related to the − oxygen storage capacity of the material (the energy stored/released experimental oxygen partial pressure should be around 10 5 atm during the reduction/oxidation is related to the reaction extent that (10 ppm O2), which corresponds to a transition temperature of about corresponds to the quantity of O 2 released/captured). As evidenced by 650–700 °C but it increases with a higher pO 2. Furthermore, the measurements and calculations, the gravimetric energy storage density calculated transition temperature depends noticeably on the composi- of mixed oxides is decreased when compared with pure Co O , tion of the system. It is thus concluded that, with a heating rate of 3 4 resulting in less storage capacity per gram of material. This observation 20 °C/min, the system does not reach equilibrium at low temperature, is in accordance with the work of Block et al. [2], who stated that both most likely because of solid-state di ffusion limitations. Independently pure Co O and Fe O show higher enthalpies of reaction than any of the system composition, a minimum temperature threshold of about 3 4 2 3 mixture of the two. 850 °C is required to achieve the phase transition. In summary, while the Co O /CoO redox pair shows very good The effect of applying a higher temperature limit than 1050 °C was 3 4 cycling stability, the addition of iron is shown to slightly a ffect it and to also estimated, both with experimental measurements and calcula- decrease drastically the overall oxygen storage capacity. According to tions. Cycles with a maximal temperature of 1150 °C were carried out equilibrium calculations in the Co-Fe-O system, the addition of iron to on Co 3O4 with 5 mol% Fe ( Fig. SI7). While this sample behaved well Co O results in a lower amount of O exchanged during cycles and a with a maximum temperature set at 1050 °C, the heating up to 1150 °C 3 4 2 loss of the reduction reaction enthalpy, due to an incomplete conver- clearly affected its reactivity. Indeed, the sample did not regain the sion of the spinel phase at 1050 °C and to the formation of a non- whole mass on cooling (67% average instead of 86% when heating at stoichiometric monoxide. The addition of iron also increases the 1050 °C) and was sintered at the end of the run. reduction and oxidation temperatures, while slightly decreasing the Furthermore, equilibrium calculations show that applying a tem- gap in temperature between the reduction and the oxidation step. perature limit higher than 1050 °C will not increase drastically the oxygen storage capacity of mixed Co-Fe oxides, as illustrated by Fig. 8. Indeed, for Fe contents between 0% and 10%, the oxygen storage 3.2. Manganese-based mixed oxides capacity is almost constant with temperature, because the non- stoichiometry of the monoxide is stable. For higher Fe contents The calculated Mn-Fe-O phase diagram is presented in Fig. 10 , for (0.10 < x(Fe) < 0.40), a small increase of the storage capacity is pO 2 = 0.20 atm. Again, the expected behavior of the mixed oxides is evidenced at high temperature, due to a higher conversion rate of the strongly dependent on the system composition. At Fe contents above spinel phase (Fig. 3b), but Fe addition is still detrimental for the total 50%, the requested temperature to reach a full conversion of the initial storage capacity. oxide into a spinel phase is above 1100 °C. At very low Fe contents (0 As reported in Fig. 9, the measured reaction enthalpies are < x(Fe) < 0 .05), the stable phase at 1050 °C is the tetragonal spinel. For 0.05 < x(Fe) < 0.15, the transition of bixbyite to cubic spinel goes

Fig. 9. Experimental and calculated enthalpies of reaction for the Co-Fe and Mn-Fe mixed oxides. Fig. 10. Calculated Mn-Fe-O phase diagram at pO 2 = 0.20 atm. Fig. 11. TGA of Mn 2O3/Mn3O4 with addition of (a) 50 mol% Fe, (b) 40 mol% Fe, (c) 30 mol% Fe, (d) 20 mol% Fe, (e) 15 mol% Fe, and (f) 10 mol% Fe. through two consecutive two-phase zones (T-Spin+Bixb and T-Spin+C- Spin). Finally, above about 20 mol% Fe, the material transforms almost directly from the bixbyite phase to the cubic spinel. According to the model [15], the non-stoichiometry of the oxide phases should not play a signi ﬁcant role regarding oxygen storage capacity. Indeed, the bixbyite phase (Mn1−xFe x)2O3 is considered as fully stoichiometric. For the spinel phases T-Spin and C-Spin, a cationic non-stoichiometry is considered in the model, with the presence of cationic vacancies. However, as shown in Fig. SI8, the vacancies concentration at 1050 °C is at the most 1.4×10 −4 mol per mol of Mn +Fe, which leads to a negligible impact on the oxygen storage capacity.

In the composition range 0 < x(Fe) < 0.5 considered in this study, the Fig. 12. (a) Evolution of experimental Δm (%) during reduction compared to theoretical redox cycle is thus represented by: Δm (pO 2 = 0.20 atm), (b) Experimental temperatures at peak reaction rate for

manganese-based oxides in 20% O 2/Ar and comparison with the calculated temperature (Mn1−xs Fe)O x23() = 2/3(Mn 1− xs Fe)O x34() + 1/6O(g 2 ) (7) of the Bixb/Spinel transition.

At equilibrium, according to the model, the theoretical amount of value for each cycle remains close both to the maximum Δm value and O2 exchanged between 750 °C and 1050 °C remains thus the same in to the equilibrium value, which denotes negligible deactivation during the 0 < x(Fe) < 0.5 composition range. redox cycling. The addition of iron to Mn 2O3 thus increases the re- The TGA, presented in Fig. 11 , shows that the sample with the oxidation yield of the material and enhances the cycling stability, as composition x(Fe) = 0.10 is unable to regain its full mass during recently observed by Carrillo et al. [11]. oxidation, as the mass lost during the first reduction step is not Experimental measurements clearly indicate that the minimum Fe recovered during the re-oxidation step, similarly to the case of pure content necessary to improve the TES properties of mixed Mn-Fe Mn 2O3 (TGA shown in Fig. SI9). Conversely, the other compositions, oxides lies between 10 and 15 mol% Fe. The phase diagram indicates from 15 mol% Fe to 50 mol% Fe, are regaining their lost mass in a that this corresponds to the formation of the cubic spinel phase only, complete reversible way. XRD analysis of Mn-Fe samples after TGA compared to lower Fe contents, where the tetragonal spinel cannot be was performed (Fig. SI2). The XRD pattern of Mn 2O3 with 10 mol% Fe reversibly oxidized during cooling. The addition of 20 mol% Fe in fi is identi ed as a mixture of tetragonal spinel structure and Fe 2O3 (Fig. Mn 2O3 was previously mentioned as the optimal composition for SI2e), which con firms the poor re-oxidation yield of the cycled obtaining the highest enthalpy and most stable re-oxidation yields material. Conversely, the XRD pattern of Mn 2O3 with 20, 30, 40 and [11]. However, the authors also obtained full conversion for all their fi – 50 mol% Fe after TGA is identi ed as (Mn 1−xFe x)2O3 (Fig. SI2a d) with tested samples regardless of the Fe content (especially below 10% Fe), Fe 2O3 visible on the sample with the highest amount of Fe added ( Fig. which contrasts sharply with the results of the present study in which

SI2a), which denotes the complete re-oxidation of the samples after pure Mn 2O3 and Mn 2O3 sample with 10 mol% Fe were not able to fully TGA. SEM characterization of the cycled materials compared with the recover the O2 lost mass. More recently, an in-depth kinetic and fresh ones reveals sintering regardless of the amount of Fe ( Fig. SI10), mechanistic study focused on the 20 mol% Fe composition was carried which does not alter the cycling ability of Mn-Fe mixed oxides. out by the same group [23]. It is stated that the cationic distribution in Sintering is thus not the cause of the reactivity loss in the case of pure the spinel structures (cubic and tetragonal) might be su fficiently Mn 2O3 and Mn 2O3 mixed with 10 mol% Fe. di fferent to explain the strong variation in the kinetics of the oxidation Experimental and theoretical mass variations, Δm, were compared. reaction. The use of the thermodynamic model proposed by Kang and As illustrated in Fig. 12 a, all the samples feature the same Δm during Jung [15] brings support to this explanation. Indeed, as illustrated in the first reduction, in good accordance with the equilibrium calcula- Fig. 13 a, the calculated cationic distribution evidences that the amount 2+ tions summarized by (Eq. (7)). However, pure Mn 2O3 shows cycling of Mn on octahedral sites of the spinel strongly increases in the cubic stability issues as already evidenced by [8]. This phenomenon is spinel. Furthermore, it has been shown that Mn 2+ on octahedral sites is re flected by the drop of the experimental Δm during 2nd and 3rd easier to oxidize than Mn 2+ on tetragonal sites [24]. This di fference cycles, showing a decrease of redox activity with redox cycling. might very well explain the reason why the cycling properties of the Similarly, Mn 2O3 mixed with 10 mol% Fe also loses cycling stability, cubic spinel are much better than those of the tetragonal spinel. regaining only 31% of its lost mass ( Fig. 11 f). From x(Fe) = 0.15 to Another di fference between the two spinel structures is that, in the 0.50, a great improvement is evidenced, since the experimental Δm oxide and similar to the Δm measured with 20% O 2. Again, the reduction temperature increases with the O2 partial pressure (Fig. SI11). The onset temperature for reduction under inert atmosphere rises with the amount of Fe added to manganese oxide, from about

800 °C for pure Mn 2O3 up to 903 °C for x(Fe) = 0.5 ( Fig. SI11 ). As illustrated by the phase diagram ( Fig. SI12) calculated at low pO 2 (10−5 atm), decreasing the oxygen content results in enhancing the tetragonal spinel stability, which is likely to be detrimental to the cycling stability of the materials. As for the energy storage density, the measured enthalpy for the

ﬁrst reduction step with full conversion of pure Mn 2O3 (148 kJ/kg) is smaller than theoretical estimations of 190.1 kJ/kg at turning temperature (ΔG° = 0) of 915 °C [10]. An average of 187.7 kJ/kg is measured for the samples with compositions between 20 and 50 mol% Fe. Increasing the amount of Fe thus slightly improves the energy storage capacity of the material at low Fe contents while it remains unchanged above ~20 mol% Fe ( Fig. 9). Accordingly, the thermody-

namic calculations indicate that an increase of ΔrH is expected when the Fe content increases from 0 mol% (207 kJ/kg) to 20 mol% (300 kJ/ kg), and is stable for higher Fe contents. In summary, the addition of iron to manganese oxide enhances considerably the cycling stability of the material by improving the re- oxidation yield thanks to the formation of a reactive cubic spinel phase.

Also, the higher the amount of Fe added to Mn 2O3, the higher the reaction temperatures. In addition, the gap in temperature between the reduction and the oxidation decreases with higher iron content. The

low re-oxidation yield observed for pure Mn 2O3 and Mn 2O3 with 10 mol% Fe is attributed to the low reactivity of the tetragonal spinel

phase. The addition of Fe to Mn 2O3 becomes effective for the enhancement of the cycling stability above ~15 mol% Fe, with the suppression of any transition involving this tetragonal spinel phase. The markedly improved reactivity of the cubic spinel featuring reversible reactions is attributed to the increased amount of Mn 2+ cations (resulting from Mn 3+ disproportionation) on octahedral sites of the spinel. Fig. 13. (a) Evolution of the amount of Mn 2+ on octahedral sites of the spinel phases with Fe content, at 1050 °C. (b) Evolution of the oxidation number of Mn cations in the 4. Conclusion spinel phases with Fe content, at 1050 °C. The effect of Fe addition in Co and Mn-based oxides was studied cubic spinel, a disproportionation reaction takes place according to experimentally and results were compared with thermodynamic calcu- 3+ 4+ 2+ Mn = 1/2 Mn + 1/2 Mn , as illustrated in Fig. 13 b. The increased lations. The calculations were shown to be pertinent since equilibrium 2+ amount of Mn in the cubic spinel, which can be more easily oxidized is reached at least on the first reduction step of all materials considered than Mn 3+ thanks to the favored thermodynamic driving force, may in this study. The addition of Fe to Mn 2O3 was shown to be bene ficial to also explain the enhanced re-oxidation ability of the cubic spinel. This tune the temperature of redox reactions, to reduce the gap in result means that pure Mn 2O3 (that reduces into pure tetragonal temperature hysteresis between the reduction and the oxidation step, spinel) and mixed Mn-Fe oxides with Fe content below 15% cannot as well as to enhance the re-oxidation kinetics and cycling stability of be suitable candidates for achieving reversible reactions because of the the material by countering the deactivation issue of Mn 2O3. poor re-oxidation ability of the tetragonal spinel formed upon reduc- Furthermore, a Fe content of ~15 mol% added to Mn 2O3 was identified tion. as a minimum threshold for avoiding the formation of a low reactive As for the in fluence of Fe incorporation in Mn O on the reaction 2 3 Mn 3O4 tetragonal spinel phase during reduction, which is detrimental temperatures, a similar tendency as with cobalt oxide is observed. The to the reaction reversibility because of poor oxidation rate and yield. reduction and oxidation temperatures of Mn O softly increase with 2 3 Noticeably, the Mn 2O3 compounds with Fe content in the range 15 – increasing amount of added iron ( Fig. 12 b). According to TGA, an 50 mol% could be cycled between bixbyite and cubic spinel phases increase of 60 °C is noted between the onset temperatures for reduction without any reactivity losses during redox reactions. Conversely, the (Fig. SI11) of pure Mn O and Mn O with 50 mol% Fe. 2 3 2 3 incorporation of Fe to Co 3O4 shows adverse effect on the redox Concomitantly, a temperature increase of 160 °C is observed for performances since both the maximum amount of O 2 exchanged and oxidation of Mn 2O3 with 50 mol% Fe when compared to Mn 2O3 alone. the reaction enthalpy are lowered when the amount of iron added is This way, the reaction temperature can be tuned. On top of an increase increased. The increasing amount of iron added to Co 3O4 reduces the of the reaction temperature, addition of iron also reduces the tem- maximum oxygen exchange capacity during a redox cycle. The addition perature gap between the reduction and the oxidation step. A tem- of iron to Co and Mn-based oxides also results in an increase of the ff perature increase was also reported by [11], with 60 °C di erence reaction temperatures, while slightly lowering the gap in temperature between 0 and 40 mol% Fe added for the reduction, and 236 °C for the between the reduction and oxidation step, which thereby reduces the oxidation. sensible energy losses during the heating and cooling stages. The ff Regarding the e ect of the gaseous atmosphere composition, the comparison of the obtained experimental data with thermodynamic Δ experimental and theoretical m values for the reduction in pure Ar modelling proves that equilibrium calculations bring a strong support atmosphere were found stable with the addition of Fe to manganese for predicting the behavior of mixed oxides since it results in consistent phase transition temperatures and reliable oxygen storage capacities. It in CSP plants, Energy Procedia 49 (2014) 820 –829. [8] A.J. Carrillo, D.P. Serrano, P. Pizarro, J.M. Coronado, Improving the thermo- further provided new physical insights into the role of Fe incorporation chemical energy storage performance of the Mn 2O3/Mn3O4 redox couple by the on the oxidation improvement of Co and Mn-based oxides related to incorporation of iron, ChemSusChem 8 (2015) 1947 –1954. the crystallographic transformations that take place during such redox [9] C. Pagkoura, G. Karagiannakis, A. Zygogianni, S. Lorentzou, M. Kostoglou, A.G. Konstandopoulos, M. Rattenburry, W.J. Woodhead, Cobalt oxide based process depending on Fe content. With the development of new models structured bodies as redox thermochemical heat storage medium for future CSP covering large multi-components systems, it could be further used for plants, Sol. Energy 108 (2014) 146–163. selecting other transition metals to enhance the properties of mixed [10] L. André, S. Abanades, G. Flamant, Screening of thermochemical systems based on metal oxides for thermochemical energy storage application. solid-gas reversible reactions for high temperature solar thermal energy storage, Renew. Sustain. Energy Rev. 64 (2016) 703 –715. [11] A.J. Carrillo, D.P. Serrano, P. Pizarro, J.M. Coronado, Thermochemical heat fl Acknowledgments storage based on the Mn 2O3/Mn3O4 redox couple: in uence of the initial particle size on the morphological evolution and cyclability, J. Mater. Chem. A 2 (2014) 19435–19443. This study was performed in the framework of the STAGE-STE [12] A.J. Carrillo, D.P. Serrano, P. Pizarro, J.M. Coronado, Manganese oxide-based European Project (FP7, Project no. 609837). Philippe Tailhades is thermochemical energy storage: modulating temperatures of redox cycles by Fe-Cu warmly thanked for the fruitful discussion on spinel properties. Co-doping, J. Energy Storage 5 (2016) 169 –176. [13] I.H. Jung, S.A. Decterov, A.D. Pelton, H.M. Kim, Y.B. Kang, Thermodynamic evaluation and modeling of the Fe-Co-O system, Acta Mater. 52 (2004) 507 –519. Appendix A. Supplementary material [14] W.-W. Zhang, M. Chen, Thermodynamic modeling of the Co –Fe –O system, Calphad 41 (2013) 76 –88 . [15] Y.B. Kang, I.H. Jung, Thermodynamic modeling of oxide phases in the Fe-Mn-O Supplementary data associated with this article can be found in the system, J. Phys. Chem. Solids 98 (2016) 237 –246. online version at doi:10.1016/j.jssc.2017.05.015. [16] L. Kjellqvist, M. Selleby, Thermodynamic assessment of the Fe-Mn-O system, J. Phase Equilib. Di ffus. 31 (2010) 113 –134. [17] E. Beche, G. Peraudeau, V. Flaud, D. Perarnau, XPS unvestigation of (La O ) − References 2 3 1 x (CeO2)2x (ZrO2)2 compounds elaborated by solar energy, Surf. Interface Anal. 44 (2012) 1045–1050. [1] A.H. Abedin, M.A. Rosen, A critical review of thermochemical energy storage [18] J.i. Lai, K.V.P.M. Shafi, A. Ulman, N.-L. Yang, M.-H. Cui, T. Vogt, C. Estournès, systems, Open Renew. Energy J. 4 (2011) 42 –46 . Mixed iron-manganese oxide nanoparticles, Nanopart. Chem. Soc. Div. Fuel Chem. [2] T. Block, N. Knoblauch, M. Shmücker, The cobalt-oxide/iron-oxide binary system 48 (2) (2003) 729 –730. for use as high temperature thermochemical energy storage material, Thermochim. [19] C.W. Bale, E. Bélisle, P. Chartrand, S.A. Decterov, G. Eriksson, A.E. Gheribi, Acta 577 (2014) 25 –32 . K. Hack, I.H. Jung, Y.B. Kang, J. Melançon, A.D. Pelton, S. Petersen, C. Robelin, [3] C. Agra fiotis, M. Roeb, M. Schücker, C. Sattler, Exploitation of thermochemical J. Sangster, P. Spencer, M.A. Van Ende, FactSage thermochemical software and cycles based on solid oxide redox systems for thermochemical storage of solar heat. databases, 2010–2016, Calphad 54 (2016) 35 –53 . Part 2: redox oxide-coated porous ceramic structures as integrated thermochemical [20] M. Hillert, The compound energy formalism, J. Alloy. Compd. 320 (2) (2001) reactors/heat exchangers, Sol. Energy 114 (2015) 440–458. 161–176. [4] C. Agra fiotis, S. Tescari, M. Roeb, M. Schücker, C. Sattler, Exploitation of [21] B. Ehrhart, E. Coker, N. Siegel, A. Weimer, Thermochemical cycle of a mixed oxide thermochemical cycles based on solid oxide redox systems for thermochemical for augmentation of thermal energy storage in solid particles, Energy Procedia 49 storage of solar heat. Part 3: cobalt oxide monolithic porous structures as integrated (2014) 762–771. thermochemical reactors/heat exchanger, Sol. Energy 114 (2015) 459–475. [22] T. Block, M. Schmucker, Metal oxides for thermochemical energy storage: a [5] M. Neises, S. Tescari, L. De Oliveira, M. Roeb, C. Sattler, B. Wong, Solar-heated comparison of several metal oxide systems, Sol. Energy 126 (2016) 195 –207. rotary kiln for thermochemical energy storage, Sol. Energy 86 (2012) 3040 –3048. [23] A.J. Carrillo, D.P. Serrano, P. Pizarro, J.M. Coronado, Understanding redox [6] S. Tescari, C. Agra fiotis, S. Breuer, L. De Oliveira, M. Nieses-von Puttkamer, kinetics of iron-doped manganese oxides for high temperature thermochemical M. Roeb, C. Sattler, Thermochemical solar energy storage via redox oxides: energy storage, J. Phys. Chem. C 120 (2016) 27800 –27812. materials and reactor/heat exchanger concepts, Energy Procedia 49 (2014) [24] F. Agnoli, B. Albouy, P. Tailhades, A. Rousset, Manganites de Fer de Structure 1034–1043. Spinelle Déformée à Très Fort Champ Coercitif. Exemple de l ′Oxyde Lacunaire – [7] G. Karagiannakis, C. Pagkoura, A. Zygogianni, S. Lorentzou, A.G. Konstandopoulos, Mixte Mn 1.7Fe 1.3O4+ δ, C. R. Acad. Sci. 2 (1999) 525 530. Monolithic ceramic redox materials for thermochemical heat storage applications